<h3 id="english-is-not-good">English is not good！英语不好警告！</h3>我在这里分享的图片是原图，你可以拖入到你的comfyui查看tag以及其他信息。The images I'm sharing here are original files; you can drag them into your ComfyUI to view the tags and other metadata.也许类别我应该选Klein那边，但我想这个流主要还是为Anima / il 这些动漫风格的模型提供修改。Perhaps I should have selected the Klein category, but I think this workflow is primarily intended for modifying anime-style models like 'Anima / il'.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/98078e9d-75c4-4083-affa-dd9fe11e5ea0/original=true/98078e9d-75c4-4083-affa-dd9fe11e5ea0.jpeg" /><hr />2.0在新版本中，使用了flux-2-klein-9b-kv-fp8的新模型 跟一致性lora。比起之前，能更容易的出带摩尔纹质感的丝袜之类，效果也会更好些。一致性lora的加入使用，能更贴合原图的那种光影氛围感，能减少二次元底图跟过于精细的三次元质感的突兀融合。想要摩尔纹直接写“增加并增强摩尔纹质感”就行了。在使用klein过程中想要有好的效果，正确的正面提示词占比很大的程度，lora的强度数值，CFG的数值，都有一定的影响，我提供的参数并不一定是最优解，多多尝试吧，玩得开心。The new version incorporates the flux-2-klein-9b-kv-fp8 model and a Consistency LoRA. It generates realistic textures—such as moiré effects on pantyhose—more effortlessly and with higher quality. The Consistency LoRA ensures better alignment with the original lighting and mood, seamlessly blending 2D anime bases with ultra-detailed 3D textures. To trigger the moiré effect, just add "increase and enhance moiré texture" to your prompt.Getting the best results out of Klein relies heavily on well-crafted positive prompts, alongside fine-tuning LoRA weights and CFG values. My recommended parameters may not be optimal for every setup, so experimentation is highly encouraged. Enjoy generating!工作流跟模型查看附件，lora在下面。Workflow &amp; Model: see attachments. LoRA: down below.<hr />1.0这是一个平平无奇的Klein局部重绘工作流。只是相较于其他的重绘替换工作流，它或许适合专门用来重绘动漫图片的pantyhose来为其优化增加纹理或者改变材质。内含EasySize 节点，其中EasySize管理器好像找不到，所以我会另外上传。Just a plain and simple Klein inpainting workflow.However, compared to other inpainting or replacement workflows, it is perhaps specifically suited for inpainting 'pantyhose' in anime images to optimize and add texture or change materials.It includes EasySize nodes. Since the EasySize manager seems to be missing, I will upload it separately.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/059176ca-9130-4bb3-b920-8e0703704153/original=true/059176ca-9130-4bb3-b920-8e0703704153.jpeg" />实在想不出要写啥玩意还是交给ai吧，让千问简单推理一下纹理的关键词自己看着加优化就是了。当然这一块没有千问，嫌装的麻烦不装直接删掉也行，只是提供个选择。I really can't think of what to write, so I'll just leave it to the AI. Let Qwen do some simple reasoning for the texture keywords, and I'll just look at them and add optimizations as needed.Of course, if you don't have Qwen and find it too troublesome to install, you can just delete it directly. It’s just provided as an option.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fe829815-3ad5-4aff-9213-dfdc54a393ab/original=true/fe829815-3ad5-4aff-9213-dfdc54a393ab.jpeg" />普普通通的上传图片添加蒙版，这里决定最终图像大小，你可以改成你上传图片的长宽之中的最长边，当然不推荐改太大，时间久又卡，具体请看个人配置自己尝试。Upload the image and add a mask. This step determines the final output resolution; you can set it to the longest edge of your uploaded image's dimensions.However, I don't recommend setting it too high, as it can be slow and may cause lag. Please try it out based on your own hardware configuration.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4f4ab486-eb2c-4bbf-a548-7e0886f979f0/original=true/4f4ab486-eb2c-4bbf-a548-7e0886f979f0.jpeg" />关于工作流内Detail Daemon这个插件，简单来说如果你觉得细节-噪点这些多了，可以把它改为负数，-0.3/-0.5之类，现在的这个参数只是目前个人实验下来觉得还算可以的一个参数。具体的调整参数可以查看该插件在仓库那边的文档，按大家喜好调整。哦对了 左边这个只是曲线预览表 右边的才是参数使用的主体。Regarding the 'Detail Daemon' plugin in the workflow: simply put, if you feel there is too much detail or noise, you can set it to a negative value, such as -0.3 or -0.5. The current parameters are just what I’ve found to work reasonably well through my personal experiments.For specific parameter adjustments, please refer to the plugin's documentation and adjust them according to your personal preference.Oh, and one more thing: the table on the left is just a curve preview; the one on the right is the actual component where you apply the parameters.<hr />tag：黑色裤袜,black pantyhose,为这条裤袜的整体增加高清细腻纹理材质，表面哑光，呈现高弹性的尼龙丝袜的细腻质感，增强材质质感。符合原图的光影氛围。/什么颜色改成什么颜色就是了，目前试出来哑光这块就这个比较满意。像其他材质，网袜这块我记得关键点的tag是“编织纹理‘’。高光泽亮度的油光袜、棉袜之类的还没怎么去试出比较满意的tag。大伙要是试出了什么好用有效果的tag也欢迎留言分享一下。<hr />其他.....好像也没什么好说的了。推荐使用下面的两个Klein pantyhose类lora来稳定一下脚趾那块的袜底，让它不要是透明或者说有间隙的。一个泛用点，一个专门是black pantyhsoe这块，具体怎样，自己尝试吧。Klein的版本推荐用下面评论中的新版本。I recommend using the two Klein pantyhose LoRAs listed below to stabilize the sole of the stocking near the toes, ensuring it isn't transparent or showing gaps. One is for general use, and the other is specialized for black pantyhose. As for how they perform, give them a try yourself.I recommend using the new version of Klein mentioned in the comments below.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/df8eb387-31a7-4d28-b5d0-f7ac7d6caf65/original=true/df8eb387-31a7-4d28-b5d0-f7ac7d6caf65.jpeg" />

Klein_anime_pantyhose_局部重绘优化材质-Klein_anima_pantyhose_Refine Texture via Inpainting

在新版本中，使用了flux-2-klein-9b-kv-fp8的新模型 跟一致性lora。比起之前，能更容易的出带摩尔纹质感的丝袜之类，效果也会更好些。一致性lora的加入使用，能更贴合原图的那种光影氛围感，能减少二次元底图跟过于精细的三次元质感的突兀融合。想要摩尔纹直接写“增加并增强摩尔纹质感”就行了。在使用klein过程中想要有好的效果，正确的正面提示词占比很大的程度，lora的强度数值，CFG的数值，都有一定的影响，我提供的参数并不一定是最优解，多多尝试吧，玩得开心。The new version incorporates the flux-2-klein-9b-kv-fp8 model and a Consistency LoRA. It generates realistic textures—such as moiré effects on pantyhose—more effortlessly and with higher quality. The Consistency LoRA ensures better alignment with the original lighting and mood, seamlessly blending 2D anime bases with ultra-detailed 3D textures. To trigger the moiré effect, just add "increase and enhance moiré texture" to your prompt.Getting the best results out of Klein relies heavily on well-crafted positive prompts, alongside fine-tuning LoRA weights and CFG values. My recommended parameters may not be optimal for every setup, so experimentation is highly encouraged. Enjoy generating!

Files with SHA256: 1f71115e7a3713faec3eacdda83e8a0842ba9de5c9badf71f7b58aa7abf3c30c